MPLS Demo: Some Answers, Some MIAs

As more and more carriers begin deploying Multiprotocol Label Switching (MPLS) in their networks, interoperability among different vendors’ products is critical. With this in mind, the world looked toward last week's interoperability demonstrations at the MPLS World Congress in Paris for evidence of progress.

A total of 13 vendors showed up last week. Participating equipment vendors included routing vendors Alcatel SA (NYSE: ALA; Paris: CGEP:PA), Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7), Cisco Systems Inc. (Nasdaq: CSCO), Redback Networks Inc. (Nasdaq: RBAK), and Riverstone Networks Inc. (Nasdaq: RSTN), along with Nortel Networks Corp. (NYSE/Toronto: NT) and testing gear vendors, Agilent Technologies Inc. (NYSE: A), Ixia (Nasdaq: XXIA), and Spirent plc (NYSE: SPM; London: SPT).

The tests, sponsored by the MPLS Forum and European Advanced Networking Test Center AG (EANTC), were billed as the first public scaleability tests designed to demonstrate the ability of multi-vendor MPLS VPN technology and equipment to meet the requirements of large service provider networks.

The first test of these technologies took place last June at the Supercomm tradeshow in Atlanta, and the third is scheduled for June 2003 at Supercomm.

So what type of progress has been made? Vendors showed basic interoperability, but they also demonstrated that Layer 2 Virtual Private Networks (VPNs), based on the Internet Engineering Task Force (IETF) Martini draft, and Layer 3 Virtual Private Networks (VPNs), using RFC 2547, could successfully scale to real-world limits, says Carsten Rossenhövel, managing director of research & manufacturer testing for EANTC. It was also the first interoperability demonstration to test both the facility backup and detour backup methods for implementing MPLS Fast Reroute.

“This wasn’t a bake-off in any way,” says Rossenhövel. “We talked to service providers and wrote test specifications based on their requirements. In many cases, the vendors participating could have exceeded these goals.”

But eyebrows may be raised by the point-to-point Layer 2 VPN test, in which only three routing vendors participated -- Cisco, Redback, and Riverstone. All three were able to reach the goal of 200 Ethernet VLAN tunnels.

The Layer 3 VPN demo tested the scaleability of Border Gateway Protocol-based VPNs. More than two-thirds of the seven participating vendors were able to achieve the test goal of 255 VPNs among 10 edge routers, each injecting between 10 and 1,000 routes into each VPN tunnel. Alcatel, Cisco, Data Connection Ltd. (DCL), NetPlane Systems Inc., Nortel, Redback, and Riverstone all participated.

Five vendors participated in the MPLS Fast Reroute portion of the test, demonstrating for the first time in public interoperability using the detour method. Alcatel, Avici, Data Connection, and NetPlane all support this feature on their gear, which provides one-to-one link protection and allows each link to have its own label switched path (LSP) for failover. Avici and Cisco also demonstrated the facility method, which uses one LSP for failover.

Rossenhövel says that a few slight problems occurred during the MPLS Fast Reroute test, but he emphasized that all the vendors were able to demonstration sub-50ms recovery time.

“Given that this was the first public interoperability test for detour and the second for facility backup, the test results were promising,” he wrote in the white paper detailing the testing results. “They helped to identify interoperability issues, mainly caused by not well-defined draft standards.”

While the test was deemed an overall success, it was hard not to notice that key vendors were missing from some of the demonstrations. In particular, routing and switching vendors, CoSine Communications Inc. (Nasdaq: COSN), Juniper Networks Inc. (Nasdaq: JNPR), Extreme Networks Inc. (Nasdaq: EXTR), and Foundry Networks Inc. (Nasdaq: FDRY) -- all of which have participated in other interoperability events hosted by other organizations -- were absent from this event (see MPLS Vendors Demo Fast Reroute).

It seems that for many of these vendors, resources for interoperability testing has become scarce.

“The decision not to participate was one of resources and priorities,” wrote Susan Ursch, a Juniper spokesperson in an email today. “At this year's event Juniper chose to focus its attention on demonstrating the first ever VPLS BGP/MPLS implementation.”

Foundry, which participated in about eight MPLS interoperability events in 2002, says it also chose not to participate because of limited resources.

Extreme wouldn’t give a specific reason for not participating. The company pointed out it has demonstrated MPLS interoperability with Isocore, with InteropNet Labs at the Spring and Fall N+I, at the University of New Hampshire InterOperability Laboratory (IOL), and with the MPLS Forum and Metro Ethernet Forum.

Others like CoSine seemed to think that the MPLS World Congress event was not worth their investment of time or money.

“We do not believe that these kinds of demonstrations buy much in the way of customer traction,” says Jay Nichols, a spokesperson for CoSine.

Steven Garrison, director of marketing for Riverstone, whose company also participated in many of these other events, says he sees a lot of value in participating in as many events as possible. He says that interoperability demonstrations are not a number one priority as far as marketing strategy goes, but that they provide important strategic information on competitors. He says over the last couple of years that he has seen participation decline.

“Companies are being choosier about the events they participate in,” he says. “But the ones that are really serious about a technology show up and the posers don’t.”

— Marguerite Reardon, Senior Editor, Light Reading
Page 1 / 2   >   >>
Hanover_Fist 12/5/2012 | 12:40:25 AM
re: MPLS Demo: Some Answers, Some MIAs "Foundry, which participated in about eight MPLS interoperability events in 2002, says it chose not to participate also because of limited resources."
Foundry's ability to craft fiction gets better and better with every LightReading article!

I'd REALLY be interested in their NAMING the 8 different MPLS events they attended in 2002!

They can't because there weren't that many public MPLS events.

Oh, they must be talking about the number of in house MPLS demonstrations they showed to visiting MPLS component vendors...that's it....

Yeah, and I hear that due to California state budget constraints, the Golden Gate Bridge is for sale again (a true, once in a life time revenue making opportunity)!
skeptic 12/5/2012 | 12:40:25 AM
re: MPLS Demo: Some Answers, Some MIAs It seems that for many of these vendors, resources for interoperability testing has become scarce.
No. Its more that wasting resources on events
like this in paris is something few can do these
days. The truth is that very little is
accomplished in these sorts of events in the
way of pratical interoprability.

The MPLS forum and the MPLS world congress count
for very little overall.
metroman 12/5/2012 | 12:40:13 AM
re: MPLS Demo: Some Answers, Some MIAs I suspect that Foundry looked at the test criteria and saw that they needed to create 200 L2 Ethernet Martini Tunnels and realised that they would fail the tests. They have a software based Ethernet Martini implementation which is really only for demo purposes.

lastofthebohicans 12/5/2012 | 12:40:09 AM
re: MPLS Demo: Some Answers, Some MIAs A solution looking for a problem. The bandwidth
glut has kept MPLS in the core, and will not
make it out to the edge for 2 years in any
serious deployments; most equipment vendors don't
have the money to spend on it in any event,
as reported in the article

Well, wait a minute, there's GMPLS for
optical networks ... **sip** ahh **chug** ahh ...
keep on drinkin' that mpls koolaid


metroman 12/5/2012 | 12:40:08 AM
re: MPLS Demo: Some Answers, Some MIAs lotb

I think you nearly have it but not quite. Sure MPLS as a core traffic engineering technology is a waste of time when you are underutilising your capacity. At the edge is where there is limited capacity and where fibre is becoming very cheap. If you look at the L2 VPN (in my view a point to point service is not a VPN but that is just semantics) tests that were done they are looking at doing 2 things with MPLS, firstly, replace leased line technology with point to point Ethernet based VPNs but also make more use of cheap fibre. This should also allow for the integration of broadband, backhauled over Ethernet rather than ATM and then across MPLS enabled cores. The core might then look like it might need engineering or indeed need some new optical harware GMPLS enabled or not.

GMPLS and BGP VPNs are a solution looking for a problem but Ethernet VPNs are a solution to a very real problem faced by people like me who need a great deal of bandwidth and don't want to pay SONET prices or be restriced by SONET bandwidth increments. I also don't want my operator running my IP network.

It does not solve all problems but it is a simple solution in a market that desperatly needs revenue generating, innovative thinking.

ASSumeNothing 12/5/2012 | 12:40:05 AM
re: MPLS Demo: Some Answers, Some MIAs "people like me who need a great deal of bandwidth and don't want to pay SONET prices or be restriced by SONET bandwidth increments. I also don't want my operator running my IP network.

Metroman, SONET standards include the ability to switch even at DS1 increments (AKA VT1.5). However, most large scale SONET boxes switch at STS-1 rates (28 VT1.5's in an STS-1) and terminate OC3 and up. But if there was a box that switched at VT1.5 (DS1/T1) (which allowed you to sell services at that level while still scaling to many Gbps and even Tbps in size) solve your problem in addition to Ethernet VPN?
BobbyMax 12/5/2012 | 12:40:02 AM
re: MPLS Demo: Some Answers, Some MIAs The kind of interoperability testing presently being conducted is a big joke. These testings could take months and years in stead of days and weeks. Vendors also did not submit any document describing as to how they overcome the interoperability problems. The equipment vendors have not contributed moneys for establishing international testing consortium.

The burden is on the vendors to make their equipment interoperable. They should all relevant test documents to the testing consortium. Many vendors may be rushing to grab headlines instead of making their equipment truly interoperable.Many vendors want to hit the market regardless of interoperability results.
cube 12/5/2012 | 12:39:59 AM
re: MPLS Demo: Some Answers, Some MIAs This is an honest, if naieve, question (i.e. not flame-bait!). What is the business case for MPLS at the core? These core guys have acres of written-down ATM equipment humming away so why should they switch to MPLS? These vendors are having a fine time in Paris connecting to one another, but what do they perceive to be the "hard need" for MPLS within, say the next 18-24 months?
cross 12/5/2012 | 12:39:56 AM
re: MPLS Demo: Some Answers, Some MIAs An efficient and well-planned interop test can provide valuable technical results in a short time. This test had only a week of hot-staging, but four months of test plan preparation. When vendors arrived at EANTC, everybody knew exactly what tests to run, with which configuration, physical layer ports, etc. Only the IP address plan had ten pages to speed up testing and to minimize misunderstandings ...
If vendors would just see this as a "big joke", why would they let 30 of their best test engineers spend a week there?

Please read the white paper of the MPLS World Congress interop event ( at http://www.eantc.de/mplsworld2... ). It *DOES* detail the tests conducted, the results and the problems we noticed. The "Problems" section provides recommendations to IETF and vendors where draft standards should be clarified and/or implemented carefully.

In addition, the test plan of the event (developed by MPLS Forum members EANTC, UNH, Agilent) is an MPLS Forum interop working group document. It will eventually become part of generic interop guidelines, together with test plans from other MPLS Forum events.

To me, there is no benefit of a "big joke" interop testing, not for the vendors (do not gain credibility), not to customers (quickly notice the fake and don't believe in anything in the future) and not the test lab (loses reputation). I agree some labs/events have tried this in the past, and your posting is a direct consequence. I would like to ask you to read the details of this interop test. If you still feel the same way afterwards, I look forward to a nice flame :)

Thanks, Carsten Rossenhoevel (EANTC)

metroman 12/5/2012 | 12:39:39 AM
re: MPLS Demo: Some Answers, Some MIAs I am aware of the issues in SONET increments. If I want to buy 67Mbps then this means, as you state that I need to buy OC3. I need to be able to scale to Gbps rather than Tbps so I don't want expensive boxes.

I assume you are making a sales pitch? Who for?

Page 1 / 2   >   >>
Sign In